▌ ▐ ▐▀▌ █ █▀▌ ▐▀▌ ▌ ▐ ▐▌▐▌ ▐▀▀ ▀▌▐▌▐▀ █▀█ █▀█ █ █ █ █▀█ █ █ ▐▐▌█ █▀ ▌▐▌▐ ▌ ▐ ▌ ▐ ▐██ ███ ▌ ▐ █▄█ ▐ █ ▐██ ▄▌▐▌▐▄ (W)elcome (A)dvisees (C)alendar (P)ublications (V)ita/Bio/Contact (F)AQ (T)eaching (D)ata + Code (L)inks (B)log (I)mages + Photos Tal(k)s
My thesis committee is made up of Daniel Marcu (CS), Kevin Knight (CS), Eduard Hovy (CS), Stefan Schaal (CS), Gareth James (Statistics), Andrew McCallum (UMass).

The thesis is available in PDF or Postscript format (warning: it's a big file!). BibTeX is also availble. You can also download my defense slides in either OpenOffice format or PDF (warning: animations don't come through in PDF).

The thesis abstract is:

Natural language processing is replete with problems whose outputs are highly complex and structured. The current state-of-the-art in machine learning is not yet sufficiently general to be applied to general problems in NLP. In this thesis, I present Searn (for "search-learn"), an approach to learning for structured outputs that is applicable to the wide variety of problems encountered in natural language (and, hopefully, to problems in other domains, such as vision and biology). To demonstrate Searn's general applicability, I present applications in such diverse areas as automatic document summarization and entity detection and tracking. In these applications, Searn is empirically shown to achieve state-of-the-art performance.

Searn is based on an integration of learning and search. This contrasts with standard approaches that define a model, learn parameters for that model, and then use the model and the learned parameters to produce new outputs. In most NLP problems, the "produce new outputs" step includes an intractable computation. One must therefore employ a heuristic search function for the production step. Instead of shying away from search, Searn attacks it head on and considers structured prediction to be defined by a search problem. The corresponding learning problem is then made natural: learn parameters so that search succeeds.

The two application domains I study most closely in this thesis are entity detection and tracking (EDT) and automatic document summarization. EDT is the problem of finding all references to people, places and organizations in a document and identifying their relationships. Summarization is the task of producing a short summary for either a single document or for a collection of documents. These problems exhibit complex structure that cannot be captured and exploited using previously proposed structured prediction algorithms. By applying Searn to these problems, I am able to learn models that benefit from highly complex, non-local features. Such features would not be available to structured prediction algorithm that require model tractability. These improvements lead to state-of-the-art performance on standardized data sets with low computational overhead.

Searn operates by transforming structured prediction problems into a collection of classification problems, to which any standard binary classifier may be applied (for instance, a support vector machine or decision tree). In fact, Searn represents a family of structured prediction algorithms depending on the classifier and search space used. From a theoretical perspective, Searn satisfies a strong fundamental performance guarantee: given a good classification algorithm, Searn yields a good structured prediction algorithm. Such theoretical results are possible for other structured prediction only when the underlying model is tractable. For Searn, I am able to state strong results that are independent of the size or tractability of the search space. This provides theoretical justification for integrating search with learning.



credits: design and font inspired by Seth Able's LoRD, some images converted to ANSI using ManyTools, original drawing of me by anonymous.
last updated on twenty seven march, two thousand seventeen.